• State Resolved
  • Locked Locked
  • Replies 5 replies
  • Answers 2 answers
  • Subscribers 32 subscribers
  • Views 610 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

SN75DP130 ADDR_EQ

Andrew Holme
Prodigy 140 points

March 2015 datasheet table 3 shows default "high-range EQ gain" of AEQ(L0) = 15dB at 2.7 GHz when ADDR_EQ=VIH and EQ_I2C_ENABLE=0.

Tables 11 through 18 show maximum EQ setting of 9dB for HBR when EQ_I2C_ENABLE=1.

Table 3 indicates that ADDR_EQ is X don't care when EQ_I2C_ENABLE=1.

Is "high-range" EQ only available as default?  Does ADDR_EQ have no effect when EQ_I2C_ENABLE=1?

What is default "high-range" EQ at 5.4Gb HBR2?

Please confirm / clarify / advise.

Thanks.

  • 0
    TI__Mastermind 23815 points
    Hi Andrew,

    On default configuration Link training is on, so, EQ gain depends on pre-emphasis level (L0, L1, L2, L3), This level is selected on link training.
    The 15bB gain you talk about is not fixed, it is the maximum possible gain for pre-emphasis level L0.
    So, there is no fixed default EQ gain, by default it selects the best EQ gain.

    Maximum EQ gain for HBR is 9dB, maximum gain for HBR2 is 18dB

    When EQ_I2C_ENABLE=1, value of ADDR_EQ doesn't affect EQ gain because it is being configured with I2C.

    Regards
  • 0 in reply to Moises Garcia
    Prodigy 140 points

    Thank you; but I was looking for information beyond what is stated on the datasheet.

    What is the EQ (dB) at 5.4 GHz (HBR2) when EQ_I2C_ENABLE=0 and ADDR_EQ=VIH and CAD_SNK=VIL and LINK_TRAINING is ON and the link is trained to pre-emphasis level L0?  Same question for L1, L2 and L3?

    I have a system containing two re-drivers in cascade with only a pair of Samtec ERM-8/ERF-8 connectors and short PCB traces between them.  The first and second re-drivers are wired with ADDR_EQ pulled low and high respectively.  The link is flakey at HBR2 with default I2C settings.  Pulling the second re-driver's ADDR_EQ low improves it.  Overall the link has three legs:

    AMD GPU ==> 1st re-driver (0.5m internal cable)

    1st re-driver ==> 2nd re-driver (PCB to PCB via Samtec)

    2nd re-driver ==> external sink (2m cable)

    DP 1.2 Link training is attempting to optimise 3 different links using the same pre-emphasis setting.  It is difficult to find non-default re-driver settings (e.g. fixed output PE on 1st re-driver, which implies fixed input EQ) that optimise the overall link.

    Thanks, Andrew.

  • 0 in reply to Andrew Holme
    TI__Mastermind 23815 points
    Hi Andrew,

    EQ gains is not fixed in link training mode, AEQ ( Automatic EQ ) choose the optimum EQ gain, the value you see in datasheet is the maximum posible, this works for HBR and HBR2. So, there is no default gain.

    The configuration you have is not a recommended one.
    Have you tried placing only the redriver that is close to sink?
    If you need two redrives, I recommend to fix the gain of the redriver that is close to GPU or at least have more control of it configuring it over I2C. You fix EQ gain by disabling link training.

    Regards
  • 0 in reply to Moises Garcia
    Prodigy 140 points

    Surely the re-driver simply monitors training commands on the AUX and selects one of four EQ levels according to the pre-emphasis level?

     

  • 0 in reply to Andrew Holme
    TI__Mastermind 23815 points
    Hi Andrew,

    For path between DP130 and sink.
    From link training(Monitoring AUX), source selects one of four pre-emphasis levels, DP130 just mimics it.

    For path between source and DP130
    AEQ adapts depending on link loss, this value is on the range defined by pre-empahsis level.

    Regards